Human Control
94 / 100
- Human presence
- High
- Actor continuity
- High
- Behaviour consistency
- High
- AI assistance
- Low
- External intervention
- No significant signal
A risk and evidence signal, not a verdict. The relying application decides what score its policy requires.
Proof of Control
Proof of Control is the human-control verification primitive of Proof of Actor: a proposed way to establish, with a stated confidence, that a present human is controlling a specific digital interaction throughout the session, and how materially AI is participating in it.
The problem
Human in the loop is a design. Human control is a fact to establish.
"Human in the loop" describes how a process is designed: somewhere, a person is supposed to review or approve. "Human on the loop" describes supervision of an otherwise autonomous system. Both are statements about intended oversight. Neither tells a relying service whether a human actually controlled the specific interaction in front of it.
That gap matters more every month. A person can authenticate successfully and then give an AI agent control, share the session, hand the device to someone else, permit remote control, use AI assistance or delegate future actions. The account is the same. The actor is not.
Proof of Control is designed to answer the question authentication cannot: is the same human present, continuously in control, and how much is AI contributing? It is the layer that distinguishes HUMAN from HUMAN + AI from an agent operating through a human's session.
What it establishes
What Proof of Control is designed to establish
Human presence
That a human is present during the interaction, not only at login.
Actor continuity
That the same actor continues throughout the session, addressing hand-offs, substitution, remote assistance and session takeover.
Human control
That the present human is actually controlling the interaction, expressed as a confidence, not an absolute verdict.
AI participation
A graded view of AI involvement: no meaningful signal, possible, probable or declared assistance, or autonomous agent activity.
Evidence, from more than one signal
Proof of Control is designed to reason over independent evidence throughout a session. Families of evidence the architecture is being built around may include:
We publish what is established, not how. Scoring, signal weightings and detection techniques are not disclosed.
Control requirements are contextual
- Hiring assessment
- HUMAN, continuous presence, low AI assistance
- Professional writing tool
- HUMAN or HUMAN + AI, with AI use declared
- Bank changing beneficiary details
- Verified identity + current presence + explicit human intent
- Remote examination
- Same verified candidate present for the full session
Differentiation
Proof of Control and adjacent approaches
| Approach | Answers | What it adds |
|---|---|---|
| Authentication / MFA | Can this entity access this account? | Who is actually performing this action, after login. |
| Human-in-the-loop design | Is a person supposed to approve this step? | Evidence that a person actually did control this interaction. |
| Session monitoring | Is this session behaving unusually? | A positive, explainable claim about presence, continuity and control. |
| AI-text detection | Does this content look machine-written? | Graded AI participation in the interaction, not judgement of a single output. |
API Preview
Require continuous human control (API Preview)
In the proposed design, an application declares the control it requires and receives an Actor status with a Human Control Score: a risk and evidence signal its own policy interprets.
// Developer Preview - API design is evolving
import { ProofOfActor } from "@proofofactor/sdk";
const actor = await ProofOfActor.verify({
require: [
"human_presence",
"human_control",
"session_continuity"
]
});
// The relying application decides the policy
if (actor.humanControl > 0.90) {
allow();
}Where it applies
Related use cases
FAQ
Frequently asked questions
What is human control verification?
Human control verification is the process of establishing, with evidence, that a human is actually controlling a specific digital interaction, rather than an AI agent, a remote operator or another person using the same session. It goes beyond authentication, which only proves that someone could access the account.
What is the difference between human in the loop and human on the loop?
Human in the loop means a person must take part in or approve a decision before it happens. Human on the loop means a system acts autonomously while a person supervises and can intervene. Both describe oversight by design. Proof of Control is designed to provide evidence of whether a human actually controlled a particular interaction.
Is using AI assistance treated as a failure?
No. HUMAN + AI is a legitimate Actor state. Proof of Control is designed to make AI participation visible so the relying application can apply its own policy. A writing tool may welcome AI assistance; a hiring assessment may not.
What is a Human Control Score?
A potential confidence metric summarising presence, continuity, behavioural consistency, AI assistance and external intervention. It is a risk and evidence signal, never a claim that an interaction is 100% human.
Does Proof of Control mean recording users?
No. The architecture is privacy-aware: it is designed to collect the minimum information required for the requested claim, and the application receives a claim with evidence, not raw recordings. We do not publish how individual signals are evaluated.
- Every primitive returns its result inside the same Actor Attestation, so applications can combine them.
Tell us what you need to verify.
Proof of Actor is being designed with developers and design partners. Your use case shapes the API.